Every one of SEAS5’s fifty-one members, tracked storm by storm. The pressure field on route 16 is the input; this route runs bombcyclone.earth’s own cyclone tracker through it — prominent extratropical pressure minima, nearest-neighbour linking, and the Sanders–Gyakum deepening rate that defines a bomb — and counts what it finds.
The axis is months, and that is deliberate. Route 16 draws days because a daily pressure map is a real object. A daily cyclone map at seasonal lead is not: past about two weeks, one member’s low on the 14th of January is not a forecast that a storm will be there on the 14th of January — it is one draw from a distribution. So no individual track is published here as a forecast. What is published is the monthly statistic across fifty-one members: where cyclones pass, where they form, where they explode, how many, and how far apart the members are about it. That is a real seasonal quantity — the storm track itself shifts with ENSO, the NAO and the jet on exactly these timescales.
How many storms, and how much do the members disagree
Counts are normalised to a thirty-day month, and that is not cosmetic. On the raw totals February came out at 418 cyclones against January’s 462 and read as a mid-winter lull; per thirty days they are 448 and 448. The whole apparent dip was that February is three days shorter. The bomb column survives the same test and keeps its shape — explosive deepening really does climb from September into midwinter, from 39.45 to 50.97 per member per thirty days. The deepest column is the lowest central pressure any single member produced that month — one draw, not an expectation.
Bombs by basin
Explosive cyclogenesis per member per thirty-day month, with the [min–max] across members beside each mean. The spread is the point: where the bracket is wide the ensemble genuinely does not know what the season holds, and a single number would hide that. Read down a row, not across the column — these basins are wildly different sizes, which is why the sea area sits beside them.
And the per-area ranking is not the ranking above — it reverses. In January the Southern Ocean posts almost as many bombs as the North Pacific (22.9 against 24.5), which reads as two comparably violent oceans. Divided by the sea they happen in, the North Pacific runs 0.91 bombs per 10⁶ km² against the Southern Ocean’s 0.26 — three and a half times the rate — and the Southern Ocean turns out to be the least bomb-dense of the major basins rather than one of the most. Its raw total is large because it is large: 89 million square kilometres of sea against the Pacific box’s 27. The physics agrees with the per-area figure: midwinter bombs feed on the baroclinicity where cold continental air meets a warm western-boundary current, over the Kuroshio and the Gulf Stream, and the Southern Ocean has no such land–sea contrast to sharpen its storms.
What one degree costs, measured
The tracker is tuned on ERA5 at a quarter degree; this archive is one degree. Rather than assume the penalty, the same ERA5 field was degraded to exactly this grid and tracked with everything else held fixed. The result: central pressure survives almost intact — within 0.9 hPa even for sub-970 storms, and peak deepening within 0.03 Bergeron — because an extratropical pressure minimum is hundreds of kilometres across and a degree barely touches it. What coarsening costs is detection: this route finds about 12 % fewer bombs than a quarter-degree tracking of the same field would. That figure is a measurement, and the counts above have not been inflated to correct for it.
The trap worth naming: the tracker’s neighbourhood parameter is a cell count, not a distance. Left at its quarter-degree default it spans eleven degrees on this grid and suppresses every low within about twelve hundred kilometres of a deeper one — which halved the bomb count before it was caught. It is scaled here.
What this route does not have
No skill score, and no anomalies. Forecast only: there is no SEAS5 reforecast behind this route, so every count is ABSOLUTE and carries the model's own bias. It cannot say 'stormier than normal'. The honest reference would be ERA5 tracked identically at one degree over thirty years — not SEAS5’s own reforecast — and until that exists every count on this page carries the model’s bias in the open rather than hidden inside a baseline.
Maritime only. Detections are maritime: the ocean mask rejects the invalid sea-level reduction over high terrain, and with it genuine cyclones over land. This is maritime cyclone density. A cyclone crossing Poland is a real cyclone and it is not in these counts.
Nothing poleward of eighty degrees. On a latitude–longitude grid the local-minimum test covers a vanishing physical area near the pole, so spurious minima multiply and any density built from them explodes on cells that are almost nothing. Measured on this run, thirteen per cent of unmasked detections fell poleward of eighty degrees — over four per cent of the Earth. Blanked rather than published.
Six-hourly, because that is the finest the archive offers. 6-hourly, the finest the CDS seasonal archive offers. Tested at 0.25°, 1/3/6-hourly gave 30/38/34 bombs — inside Poisson noise, so no cadence correction is claimed.